Implementation of LM assimilation suite at MeteoSwiss

Transcription

1 Implementation of LM assimilation suite at MeteoSwiss First operational implementation of LM assimilation cycle at MeteoSwiss, run since end of October Design goals: simple robust Jean-Marie Bettems Swiss Federal Office for Meteorology and Climatology COSMO Warsaw, September 2002

2 almo assimilation cycle overview Initial conditions almo LBC Observations BUFR AOF BUFR to AOF (makeaof) Observation DB (ECMWF DB soft.) GTS

3 The observation database Based on ECMWF software suite (libemos, preproc, mars client/server, ) - Uses Empress (v.8.6) as DB software - Store quality controlled data in BUFR format - GTS decoding software included gts GTS QC BUFR user - Data quality control sofware included - Extendible to non-conventional data types (radar data, satellite images, ) - Access to data via MARS requests - Transparency/configurable access to ECMWF database - Interface to Metview - Located at Swiss Center for Scientific Computing in Manno (SCSC) - Runs on a SGI Origin Data actively sent from MeteoSwiss GTS point of presence - Actual data type content: SYNOPS, SHIPS, ANETZ DRIBUS AIREPS, AMDARS TEMPS PILOTS - Data retention: currently 90 days permanent storage planned Statistics Typical assimilation cycle of 3 hours (mean values over 1700 cycles): Synops: 4690 Aircrafts: 1460 Dribus: 30 Temps: 55 Pilots: 10

4 almo assimilation cycle overview Initial conditions almo LBC Hourly GME Most recent Vorlauf Observations BUFR AOF BUFR to AOF (makeaof) Observation DB (ECMWF DB soft.) GTS

5 almo assimilation cycle overview Model archive (UNIX file system, la-files) Ext. sources (Snow, SST, soil, vegetation, O 3 ) GRIB GRIB Twice daily Interpolated GME analysis GRIB Merge fields (fieldextra) LM_DWD snow analysis Hourly BUFR AOF GRIB Initial conditions almo Observations BUFR to AOF (makeaof) LBC Hourly GME Most recent Vorlauf Observation DB (ECMWF DB soft.) GTS

6 almo initial fields Cst. ext. Param. Atm. fields Field HSURF FR_LAND SOILTYP GZ0 UV W T PP QV QC Source (: previous almo assim. cycle) standard 06 and 18 UTC Snow Vegetation Ozone VIO3 interpolation program (gme2lm) HMO3 interpolation program (gme2lm) PLCOV interpolation program (gme2lm) LAI interpolation program (gme2lm) ROOTDP interpolation program (gme2lm) T_SNOW LM-DWD + GME analysis W_SNOW LM-DWD + GME analysis W_I LM-DWD + GME analysis QV_S interpolated GME field W_G1 interpolated GME field Soil W_G2 interpolated GME field T_S interpolated GME field T_M interpolated GME field Soil clim. T_CL interpolated GME field W_CL interpolated GME field blue: derived from climatology magenta, red: analysed at DWD

7 Timetable almo assimilation cycle ( ) DWD analysis la_dwd_18 gif_dwd_00 la_dwd_06 gif_dwd_ UTC [hour] Assimilation cycle Obs.: LBC: (hourly) (3 hourly) (hourly) (hourly) (hourly) merge: (la_dwd_18) gif_dwd_12 merge: (la_dwd_06) gif_dwd_ Forecast cycle Obs.: LBC: now (3 hourly) (hourly) now (3 hourly) (hourly) Performance: 1 assimilation cycle: elapsed time (SX5_lm_assml, 4 PUs) forecast cycle: elapsed time (SX5_lm, 12PUs)

8 Outlook Real time monitoring (e.g. map of obervation increments) Improve data quality control (e.g introduction of blacklist based on real time monitoring) More frequent update of lateral boundary conditions set Dual assimilation cycle (long cut-off for high quality / short cut-off for near real time applications) Improve assimilation algorithm (but this is another chapter )